Hester van Eeren

Value of information in crime prevention: An illustration | 3 45 | Karnon, 2015), and it can lead to different estimated model outcomes (i.e. Frederix et al., 2014). This structural uncertainty is likely to be present in our model. For example, we did not account for the severity of crimes in the model states or the elevated risk of death for adolescents in the criminal state (i.e. Chassin, Piquero, Losoya, Mansion, & Schubert, 2013; Teplin, McClelland, Abram, & Mileusnic, 2005). The uncertainty of these aspects was not represented in the current value of information analysis. Future cost-effectiveness models in the field of crime prevention should therefore carefully characterize the structural uncertainty (Haij Ali Afzali & Karnon, 2015), and account for it when possible, for example by parameterization (Bilcke et al., 2011; Haij Ali Afzali & Karnon, 2015) or model averaging (Bojke et al., 2009; Haij Ali Afzali & Karnon, 2015; Jackson, Bojke, Thompson, Claxton, & Sharples, 2011). The second additional source of uncertainty is methodological uncertainty, which relates to the analytical method chosen (Bilcke et al., 2011; Briggs, 2000). Our model also represents some methodological uncertainties, such as whether or not to include the costs of crime in the model (i.e. McCollister, French, & Fang, 2010). Here, three methodological uncertainties in our model are discussed in more detail. These uncertainties could be resolved through, for example, formulating guidelines (i.e. Bilcke et al., 2011; Bojke et al., 2009) to model cost-effectiveness research in the field of crime prevention. The first methodological uncertainty concerns the societal perspective used in the model, which means that we included the costs and effects relevant to society. When considering this perspective in health care, the focus is merely on the patient, whereas this will be different in the area of crime prevention and treatment (i.e. van Zutphen, Goderie, & Janssen, 2014). In this study, we already included the direct and indirect costs of one parent, as well as direct non medical costs of the adolescent, such as the costs of contact with the police. Other costs that we did not account for, but are nevertheless relevant in the field of crime prevention are: the effect of the intervention reflected in both costs and effects, such as increased wellbeing and reduced productivity losses (i.e. van Zutphen et al., 2014), in regard to family members (e.g. parents or siblings of the adolescents). Further additional categories are reduced victim costs and increased victim wellbeing, the reduction of the number of out-of-home placements, the reduction of the costs of committed crimes to society, reduced costs of avoided crimes to society and the value of reduced fear of crime (i.e. Cohen & Piquero, 2009; van Zutphen et al., 2014). The second methodological uncertainty deals with the WTP value for a CAFY. Although we used the WTP values of Cohen et al. (Cohen & Piquero, 2009; Cohen et al., 2010) to illustrate the use of WTP in crime prevention, WTP to prevent a crime like burglary is definitely not equal to WTP per CAFY. Therefore, it is important to carefully estimate the WTP value per CAFY. In this study, for example, we could have weighted the WTP values of Cohen et al. (Cohen & Piquero, 2009; Cohen et al., 2010) by the frequency of the crimes as yearly committed by the adolescents in this study, or by the number of yearly registered crimes in the Netherlands. For clarity reasons and due to a lack of more detailed data regarding the crimes committed, we chose not to use a weighted WTP value. Furthermore, for the WTP values used, it is not exactly known which components of crime, such as investigation, prosecution, witnesses, legal aid,